How It Works
The same four-step process runs behind every language, every platform, every region: Generate, Route by risk, Review, Deliver. Here's exactly what ships on brand-voice autopilot, and what a human reviewer catches before your audience ever sees it.
Why this, why now
Google Ads has phased out manual campaign-level language targeting for Search. Its AI now matches ad copy to users automatically, blending multilingual delivery into regular serving, the same shift behind AI Max and Advantage+. For an advertiser, that's less direct control, not more: there used to be a targeting setting that at least made a language mismatch visible. Now a literal translation doesn't trip a warning, it just gets served confidently to the wrong intent. And paid search was never the only channel without a warning light: no email platform flags a tone-deaf subject line, no SMS gateway flags a message that reads as corporate rather than native, no app store flags a mistranslated listing.
The pipeline
Every variant, in every language and on every channel, from a headline to a subject line to a product listing, moves through the same sequence: the only thing that changes market to market is what Step 2 flags.
First-pass variants at volume, inside each platform's field limits and your termbase. This is where AI throughput does the heavy lifting.
Every variant is scored on four signals and sorted into On-Brief, Needs Polish, or Needs Rebuild before a human ever sees it.
In-market linguists polish anything flagged Needs Polish, fully transcreate anything flagged Needs Rebuild, and never touch work that's already On-Brief.
Into your sheet, a platform-ready export, Ads Editor, your ESP, your CMS, pushed straight to whatever runs it, audit trail included.
Step 2, in detail
Four signals, weighted, checked automatically on every variant, not a black box, and configurable per client.
| Signal | What it measures | Weight |
|---|---|---|
| TM/glossary match rate | % of the string already covered by locked termbase entries or a 100% Translation Memory hit | High: a low match rate always forces at least Needs Polish |
| Language risk profile | Historical error/complaint rate for that target language + MT engine pairing | Medium |
| Claim/regulatory density | Pricing, health, financial, superlative, or comparative claims that trip platform ad policy or local law | High: any hit forces Needs Rebuild regardless of other scores |
| Channel/format sensitivity | A Discord Quest or TikTok hook carries more brand-voice risk than a Shopping feed description | Medium |
| Tier | Trigger | Who touches it |
|---|---|---|
| On-Brief | ≥ 98% TM/glossary match, low-risk language, zero claim hits, low-sensitivity channel | System only, no human touch |
| Needs Polish | 75–97% match, or medium language risk, or medium-sensitivity channel, zero claim hits | Linguist post-edit |
| Needs Rebuild | Any claim/regulatory hit, high-risk language, high-sensitivity channel, or match rate under 75% on a hook/headline | Native transcreator, full rewrite |
Integration
Nothing about this asks you to change how your media buyers work. The output lands where they already look.
Termbase, translation memory, and QA rules live in Phrase: the engine on top handles the routing decision Phrase alone doesn't make.
One schema, one export: platform, char_limit, risk_tier, status, mapped automatically into Google Ads Editor or Meta's bulk-sheet columns.
Anything over a platform's field limit is blocked from export, not just flagged: zero truncated headlines reaching a live ad account.
A localized ad is only half the job: we map delivery back to the localized landing page and, for eCommerce clients, the Google Merchant Center feed, so price and inventory stay in sync past the click.
Beyond paid media
Paid search and paid social are where the CPA math is most visible, but the same Generate, Route by risk, Review, Deliver pipeline produces language-accurate copy for three more surfaces every performance team eventually has to solve for, part of eight marketing channels the engine covers end to end; see the full breakdown on Solutions.
Multilingual SEO copy and the answer-engine content that gets cited by ChatGPT, Gemini, and Google's AI Overviews, checked against the same glossary and claim-density rules as your paid ad copy, so your organic and paid message never contradict each other market to market.
Community posts and captions that need the same local slang and cultural read as a paid hook, without the paid-media budget behind them.
Localized titles, keywords, and screenshots for app store discovery: the same discipline as a Search headline, just a different field limit and a different algorithm.
We're the operational infrastructure underneath your existing martech and media stack, not a replacement for it: region-specific intelligence, bidding signals, content, and creator vetting, layered in without disrupting the account strategy you already own. We don't replace your agency of record, and we don't hand your creative directors a chaotic web of freelancers to manage. We make the multilingual layer underneath actually perform, one brand voice intact across every market.
The alternative
Traditional LSPs sell volume and legal-document compliance. Specialized studio LSPs sell in-game text and audio. Neither prices, staffs, or turns around against a marketing performance metric, on any channel.
| Capability | Inteprit | Traditional LSPs | Specialized studio LSPs |
|---|---|---|---|
| Core focus | Multilingual marketing performance, across paid, organic, and owned channels | Document and legal compliance | In-game immersion and audio |
| Pricing metric | Flat campaign bundles / retainers | Per-word commoditized rate | Per-word or hourly studio rate |
| Turnaround | Same-day / real-time iteration | 3–5 business days | Fixed production schedules |
| Tech integration | Direct sync with Phrase TMS, ad feeds, your ESP, and your CMS | Legacy API or manual file uploads | Custom proprietary game engines |
| Primary success metric | CTR, CPC, open rate, conversion lift | Accuracy, total word output | Narrative flow, player retention |
Send us your lowest-performing international asset. We'll run it through the same pipeline above and hand back exactly where intent and language fall out of step, free.
Request a Content Intent Audit Or talk it through first, schedule a call →